spacy-loggers

What it is and what it does

spacy-loggers is a companion package to spaCy that decouples experiment tracking and monitoring from the core library. Starting with spaCy v3.2, loggers were moved into this separate package so they can be updated independently. It provides integration layers for five external logging platforms—Weights & Biases, MLflow, ClearML, PyTorch, and CuPy—plus utility loggers for interoperating between them.

You use it by installing it alongside spaCy (it often installs automatically) and then configuring which logger to use in your spaCy training config file. Each logger sends training metrics, model checkpoints, and system information to its respective dashboard or tracking service. The package is designed for machine learning practitioners who want to monitor spaCy model training without adding bloat to the core library.

Use it for:

• Track spaCy NLP model training metrics and model artifacts in Weights & Biases dashboards.
• Log training runs and model performance to MLflow for experiment comparison and reproducibility.
• Monitor training progress across multiple spaCy jobs using ClearML's centralized task management.
• Chain multiple loggers together to send training data to multiple platforms simultaneously.
• Exclude sensitive paths or configuration values from logged configs before uploading to external services.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides logging integrations for spaCy model training, supporting Weights & Biases, MLflow, ClearML, PyTorch, and CuPy as alternate loggers decoupled from the core library.

Yes, if you use spaCy for model training and want to integrate with external experiment tracking platforms. It installs automatically with spaCy v3.2+ and has no runtime dependencies, so there is minimal friction. The dormant maintenance status is not a concern for a stable integration layer, but verify that your chosen logger (WandbLogger, MLflowLogger, etc.) is compatible with your current spaCy and external service versions before relying on it for production workflows.

Install

spacy-loggers on PyPI

pip install spacy-loggers

uv add spacy-loggers

poetry add spacy-loggers

Installing spacy-loggers

Before you install

Low install friction with no runtime dependencies. Dormant maintenance status (last commit 2023-11-03, 1068 days since release) suggests stability but limited active development.

License in practice

MIT license permits free use, modification, and distribution with minimal restrictions, making it suitable for both open-source and commercial projects.

Quickstart

pip install spacy-loggers

# In spaCy training config:
[training.logger]
@loggers = "spacy.WandbLogger.v5"
project_name = "my_project"

# or for MLflow:
[training.logger]
@loggers = "spacy.MLflowLogger.v2"
experiment_id = "1"

Requires the corresponding external logging service (wandb, mlflow, clearml, etc.) to be installed and configured separately; spacy-loggers itself is only the integration layer.

Verify before relying

• Whether all five loggers (Weights & Biases, MLflow, ClearML, PyTorch, CuPy) are actively maintained or if some are deprecated.
• Current compatibility with recent spaCy versions beyond v3.2.
• Performance impact of logging during large-scale training operations.

Package facts

Evidence: spacy_loggers-1.0.5-py3-none-any.whl

Tags

More Monitoring packages